Selmer varieties Minhyong Kim January, 2009 Posteh 1 Option 1: Finish now. Option 2: Proeed with leture. 2 0. `Philosophy' I. Non-abelian desent II. Diophantine niteness III. Preliminary remarks on non-abelian duality IV. Expliit formulas. 3 0. `Philosophy' Objets of study: Various Hom(Y; X ), say for X and Y shemes; ! X and Y anabelian shemes (Grothendiek); .. . ! X a generially smooth urve over Y = Spe(Z S ), where Z S is the ring of S integers for some nite set S of primes. The last ase is the Diophantine geometry of urves: the study of its S -integral points.a a When X is projetive, an S-integral point is the same as a rational point. 4 Usual approah emphasizes the dierenes between three ases: genus zero, genus one, and genus 2. However, we wish to fous on the parallels, espeially between (E; e); a ompat urve of genus one equipped with an integral point, and (X; b); a hyperboli urvea equipped with an integral point. a That C) is, X ( has a hyperboli metri. Equivalently, X is genus zero minus at least three points, genus one minus at least one point, or genus 5 2 . I. Non-abelian desent (E; e) ellipti urve over Z S . p prime not in S . T = S [ fpg. G := Gal(Q =Q ). Kummer theory provides an injetion E (Z S )=pn E (Z S ),!H 1 (GT ; E [P n ℄): BSD onjetured an isomorphism E (Z S ) Z p ' Hf1 (G; Tp (E )) where Tp (E ) := lim E [pn ℄ is the p-adi Tate module of E and the subsript f refers to loal `Selmer' onditions. 6 When X=Z S is a smooth hyperboli urve and b 2 X (Z S ), analogue of above onstrution is 1 Z p )) X (Z S ) ! Hf (G; H1et (X; using the p-adi étale homology Z p ) := et;p (X; b)ab H1et (X; 1 of X := X Spe(Q ) Spe(Q ). Several dierent desriptions of this map. 7 But in any ase, it fators through the Jaobian X (Z S )!J (Z S )!Hf1 (G; Tp J ) using the isomorphism Z p ) ' Tp J; H1et (X; where the rst map is the Albanese map x 7! [x℄ [b℄ and the seond is again provided Kummer theory on the abelian variety J . Consequently, diult to disentangle X (Z S ) from J (Z S ). Eorts of Weil, Mumford, Vojta. 8 The theory of Selmer varieties renes this to a tower: Hf1 (G; U4 ) - ? Hf1 (G; U3 ) ? - Hf1(G; U2) ? - 1 3 4 - .. . .. . X (Z S ) 2 1 Hf (G; U1 )= Hf1 (G; Tp J Q p ) where the system fUn g is the Q p -unipotent étale fundamental group b) of X . 1u;Q (X; p 9 Brief remarks on the onstrutions. 1. The étale site of X denes a ategory Q p) Un(X; of loally onstant unipotent Q p -sheaves on X . A sheaf V is unipotent if it an be onstruted using suessive extensions by the onstant sheaf [Q p ℄X . 2. We have a ber funtor Q p )!VetQ Fb : Un(X; p that assoiates to a sheaf V its stalk Vb . Then U := Aut (Fb ); the tensor-ompatible automorphisms of the funtor. U is a pro-algebrai pro-unipotent group over Q p . 10 3. U = U1 U2 U3 is the desending entral series of U , and Un = U n+1 nU are the assoiated quotients. There is an identiation Q p) = V U1 = H1et (X; := Tp J Qp at the bottom level and exat sequenes 0!U n+1 nU n !Un !Un 1 !0 for eah n. 11 For example, for n = 2, 0! ^2 V !U2 !V !0; for ane X . When X = E n feg for an ellipti urve E , this beomes 0!Q p (1)!U2 !V !0: When X is ompat, we get 0![^2 V=Q p (1)℄!U2 !V !0; where Q p (1),! ^2 V omes from the Weil pairing. 12 4. U has a natural ation of G lifting the ation on V , and H 1 (G; Un ) denotes ontinuous Galois ohomology with values in the points of Un . For n 2, this is non-abelian ohomology, and hene, does not have the struture of a group. 5. Hf1 (G; Un ) H 1 (G; Un ) denotes a subset dened by loal `Selmer' onditions that require the lasses to be (a) unramied outside a set T = S [ fpg, where S is the set of primes of bad redution; (b) and rystalline at p, a ondition oming from p-adi Hodge theory. 13 6. The system !Hf1(G; Un+1)!Hf1(G; Un)!Hf1(G; Un 1)! is a pro-algebrai variety, the Selmer variety of X . That is, eah Hf1 (G; Un ) is an algebrai variety over Q p and the transition maps are algebrai. Hf1 (G; U ) = fHf1 (G; Un )g is the moduli spae of prinipal bundles for U in the étale topology of Spe(Z [1=S ℄) that are rystalline at p. If Q T denotes the maximal extension of Q unramied outside T and GT := Gal(Q T =Q ), then Hf1 (G; Un ) is naturally realized as a losed subvariety of H 1 (GT ; Un ). 14 For the latter, there are exat sequenes Æ 0!H 1 (GT ; U n+1 nU n )!H 1 (GT ; Un )!H 1 (GT ; Un 1 ) ! H 2 (GT ; U n+1 nU n ) in the sense of ber bundles, and the algebrai strutures are built up iteratively from the Q p -vetor spae struture on the H i (GT ; U n+1nU n ) and the fat that the boundary maps Æ are algebrai. (It is non-linear in general.) So the underlying arhimedean input is of Hermite-Minkowski type, leading to nite-dimensionality of the H i (GT ; U n+1 nU n ). 15 7. The map na = fn g : X (Z S ) - Hf1 (G; U ) is dened by assoiating to a point x the prinipal U -bundle P (x) = u;Q (X ; b; x) := Isom (Fb ; Fx ) 1 p of tensor-ompatible isomorphisms from Fb to Fx , that is, the Q p -pro-unipotent étale paths from b to x. For n = 1, 1 : X (Z S )!Hf1 (G; U1 ) = Hf1 (G; Tp J Q p ) redues to the map from Kummer theory. But the map n for n 2 does not fator through the Jaobian. Hene, suggests the possibility of separating the struture of X (Z S ) from that of J (Z S ). 16 8. If one restrits U to the étale site of Q p , there are loal analogues 1 na p : X (Z p )!Hf (Gp ; Un ) that an be expliitly desribed using non-abelian p-adi Hodge theory. More preisely, there is a ompatible family of isomorphisms D : Hf1 (Gp ; Un ) ' UnDR =F 0 to homogeneous spaes for quotients of the De Rham fundamental group of X Q p . U DR = 1DR (X Q p ; b) U DR lassies unipotent vetor bundles with at onnetions on X Q p , and U DR =F 0 lassies prinipal bundles for U DR with ompatible Hodge ltrations and rystalline strutures. 17 Given a rystalline prinipal bundle P = Spe(P ) for U , D(P ) = Spe([P Br ℄G ); p where Br is Fontaine's ring of p-adi periods. This is a prinipal U DR bundle. The two onstrutions t into a diagram X (Z p ) - Hf1(Gp; U ) na p d na r = r D - ? DR 0 U =F whose ommutativity redues to the assertion that 1DR (X ; b; x) Br ' 1u;Q (X ; b; x) Br : p 18 9. The map DR =F 0 na : X ( Z ) ! U p dr=r Z is desribed using p-adi iterated integrals 1 2 n of dierential forms on X , and has a highly transendental natural: For any residue disk ℄y [ X (Z p ), DR 0 na dr=r;n (℄y [) Un =F is Zariski dense for eah n and its oordinates an be desribed as onvergent power series on the disk. 19 10. The loal and global onstrutions t into a family of ommutative diagrams - X (Zp ) X (Z S ) ? H 1 (G; U f n) ? - - Hf1(Gp; Un) -D UnDR=F 0 lo p where the bottom horizontal maps are algebrai, while the vertial maps are transendental. Thus, the diult inlusion X (Z S ) X (Z p ) has been replaed by the algebrai map lop . 20 II. Diophantine Finiteness Theorem 1 Suppose lop (Hf1 (G; Un )) Hf1 (Gp ; Un ) is not Zariski dense for some n. Then X (Z S ) is nite. Theorem is a rude appliation of the methodology. Eventually would like rened desriptions of the image of the global Selmer variety, and hene, of X (Z S ) X (Z p ) by extending the method of Chabauty and Coleman and the work of Coates-Wiles, Kolyvagin, Rubin, Kato on the onjeture of Birh and Swinnerton-Dyer. 21 Idea of proof: There is a non-zero algebrai funtion - X (Zp ) X (Z S ) na n ? Hf1 (G; Un ) na p;n ? - Hf1(Gp; Un) DÆlop 96=0 ? Qp vanishing on lop [Hf1 (G; Un )℄. Hene, Æ na p;n vanishes on X (Z S ). But using the omparison with the De Rham realization, we see that this funtion is a non-vanishing onvergent power series on eah residue disk. 2 22 -Hypothesis of the theorem expeted to always hold for n suiently large, but diult to prove. For example, Bloh-Kato onjeture on surjetivity of p-adi Chern lass map, or Fontaine-Mazur onjeture on representations of geometri origin all imply the hypothesis for n >> 0. That is, Grothendiek expeted Non-abelian `niteness of niteness of X (Z S ). X' (= setion onjeture) ) Instead we have: `Higher abelian niteness of X' ) niteness of X (ZS ). 23 Can prove the hypothesis in ases where the image of G inside Z p )) is essentially abelian. That is, when Aut(H1 (X; -X has genus zero; -X = E n feg where E is an ellipti urve with omplex multipliation; -(with John Coates) X ompat of genus 2 and JX fators into abelian varieties with omplex multipliation. For example, X might be for n 4. axn + byn = z n ; In the CM ases, need to hoose p to split inside the CM elds. 24 Idea: Construt a quotient U !W !0 and a diagram - X (Z p ) X (Z S ) na n na p;n ? lo ? lo Hf1 (G; Un ) Hf1 (G; Wn ) ? - Hf1(Gp; Un) D- UnDR=F 0 p ? - Hf1(Gp; Wn) p 25 ? - WnDR=F 0 D suh that dimHf1 (G; Wn ) < dimWnDR =F 0 for n >> 0. When JX has CM, an onstrut a `polylogarithmi quotient'a W = U=[[U; U ℄; [U; U ℄℄ suh that all CM haraters i1 i2 i n appearing in W n =W n+1 have multipliity one. a The terminology is adopted by analogy with the quotient of the fundamen- tal group of P1 n f0 1 1g ; ; that provides the natural setting for the theory of polylogarithms aording to Beilinson and Deligne. 26 Elementary Iwasawa theory an be used to show that H 2 (GT ; i1 i2 i n )=0 rather generially, allowing us to ontrol dimHf1 (G; Wn ) X dimH n i=1 X dimH n i=1 1 i =W i+1 ): ( G; W f 1 (GT ; W i =W i+1 ) and show that it grows more slowly than dimWnDR =F 0 : 27 III. Preliminary remarks on non-abelian duality As far as the arithmeti of urves is onerned one major goal of the theory is to give an expliit desription of X (Z S ) X (Z p ): This should ome from a non-abelian loal-global duality together with a non-abelian expliit reiproity law.a Wish to provide a hint of this idea using a very speial situation: X = E n feg; where EQ is an ellipti urve suh that L(EQ ; 1) = 0; L0 (EQ ; 1) 6= 0: Here, we will assume that E is in fat a minimal Z -model of EQ and X = E n feg. a Of ourse both notions are highly speulative at present. 28 Choose S so that ES := E Spe(Z) Spe(Z S ) is smooth. Our assumptions imply that is nite and X(E) rankE (Z ) = rankE (Z S ) = rankE (Q ) = 1; but still diult to analyze the inlusion X (Z ) E (Z ): Some arhimedean progress using Diophantine approximation supplemented by LLL-algorithm. Here, we fous on non-arhimedean tehniques and the inlusion X (Z ) X (Z p ): 29 Pik a tangential base-point b 2 Te E and only use U2 , identied with its Lie algebra L2 . L2 has a natural graded struture L2 = L[1℄ L[2℄ suh that L[1℄ ' L1 ' V (E ) and L[2℄ ' Q p (1) is the one-dimensional enter of L2 . The group law on U2 an be thought of as a twisted operation on L2 given bya (l1 + l2 ) (l10 + l20 ) = l1 + l10 + l2 + l20 + (1=2)[l1 ; l10 ℄: The grading is ompatible with the Galois ation (in fat, with the motivi struture): g(l1 + l2 ) = g(l1 ) + g(l2 ): a This kind of struture is often referred to as a 30 Heisenberg group . If a is a ohain on GT or Gp with values in U2 , then we an write a = a1 + a2 with a1 taking values in L[1℄ and a2 taking values in L[2℄. The oyle ondition for the group law is written in terms of these omponents as da1 = 0 da2 = (1=2)a1 [ a1 ; where the up produt of two ohains and is dened by [ (g; h) = [ (g); g(h)℄: 31 We will onstrut a funtion on Hf1 (Gp ; U2 ) using seondary ohomologial operations. Let : GT !Q p be the log of the p-adi ylotomi harater and p = jGp . Given a oyle = 1 + 2 : Gp !U2 , the funtion is, in essene, 7! (; 1 ; 1 ); where the braket refers to a Massey triple produt, taking values in H 2 (Gp ; L[2℄) ' Q p : This notion omes from rational homotopy theory, and is usually dened for ohomology lasses of an assoiative dierential graded algebra A. 32 If [a℄ 2 H 1 (A); [b℄ 2 H 1 (A); [℄ 2 H 1 (A) are lasses with the property that [a℄[b℄ = 0; [b℄[℄ = 0; then we an solve the equations dx = ab; dy = b: We see then that x + ay is a oyle, dening a lass in H 2 (A). Note that the lass depends on the hoie of x and y, a dening system. Well-dened lass lives only inside H 2 (A)=[aH 1 (A) + H 1 (A)℄: 33 In our situation, the omplex of ohains on Gp with values in Q p L(1) L(2) forms an assoiative dierential graded algebra. Given a oyle = 1 + 2 : Gp !U2 ; we have [p ℄ [ [1 ℄ = 0; sine H 2 (Gp ; L(1)) = 0: Also, sine d( 22 ) = 1 [ 1 : [1 ℄ [ [1 ℄ = 0; Therefore, we an form the Massey triple produt (p ; 1 ; 1 ) 2 H 2 (Gp ; L[2℄)=[p [ H 1 (Gp ; L[1℄)) + 1 [ H 1 (Gp ; L[1℄℄: Unfortunately, zero. 34 But note that this naive Massey produt does not use the full data of or the strength of our assumptions. Firstly, part of a dening system for the Massey produt is enoded in : d( 22 ) = 1 [ 1 : Seondly, if [ ℄ 2 Hf1 (Gp ; U2 ), then [1 ℄ 2 Hf1 (Gp ; L[1℄) is in the image of the loalization map Hf1 (G; L[1℄) ' Hf1 (Gp ; L[1℄): Hene, the equation dx = [ 1 makes sense globally. 35 Key point: Using Using our restritive hypotheses, there is a global solution xglob : GT !L[1℄ to the equation Uses the niteness of dx = [ 1 : X and a generator for E(Z). 36 Proposition 2 The lass glob p 2 p ( ) := [lop (x ) [ 1 + [ ( 22 )℄ 2 H (Gp ; L[2℄) is independent of all hoies. Thus, 1 p : Hf (Gp ; U2 )!Q p is a well-dened algebrai funtion on the loal Selmer variety. Remark: The map Z p Q p ' Hf1(Gp ; L[2℄),!Hf1(Gp ; U2) ! Q p p is the log map, and hene, p is non-zero. 37 Dene (`rened Selmer variety') Hf;1 Z(G; U2 ) Hf1 (G; U2 ) to be the intersetion of the kernels of lol : Hf1 (G; U2 )!Hf1 (Gl ; U2 ) for all l 6= p. In fat, we have a ommutative diagram - X (ZS ) X (Z ) ? ? Hf;1 Z(G; U2 ) - Hf1 (G; U2 ) Joint work with A. Tamagawa. 38 Theorem 3 (loal-global duality) Hf;0 (G; U2 ) 1 lop The map ! Hf1(Gp ; U2) ! Q p p is zero. Proof is straightforward using the standard the exat sequene 0!H 2 (GT ; Q p (1),! v2T 39 H 2 (Gv ; Q p (1))!Q p !0: IV. Expliit formulas Choose a Weierstrass equation for E and let = dx=y; = xdx=y: Dene log (z ) := Z z b ; log (z ) := D2 (z ) := Z via (iterated) Coleman integration. 40 b z ; Z z b ; Corollary 4 Then the set Suppose we have a point y 2 X (Z ) of innite order. X (Z ) X (Z p ) lies inside the zero set of the analyti funtion log2 (y ))(D2 (z ) Atually, [D2 (z ) log (z ) log (z )) pÆD 1 log (y ) log (y )): Æ na DR=r;2 = Rese (vdx=y ) log (z ) log (z ) where dv = xdx=y . log2 (z )(D2 (y ) log (z ) 2 ( ) (D2 (y ) log (y ) 41 log (y ) log (y ))℄; Remarks: -In partiular, the funtion (D2 (z ) log (z ) log (z ))=(log (z ))2 is onstant on the integral points of innite order. -Parts of this onstrution generalize to ane urves X of genus g 2 whose Jaobians have Mordell-Weil rank g. -Also to ompat urves X provided rankNS (JX ) 2: ! Possibility of omputing points on urves of genus 2 with rank 2 Jaobians. 42